Arxiv:1808.07848V2 [Quant-Ph] 14 May 2019

Detecting Einstein-Podolsky-Rosen steering through entanglement detection

Debarshi Das,1, ∗ Souradeep Sasmal,1, † and Sovik Roy2, ‡ 1Centre for Astroparticle Physics and Space Science (CAPSS), Bose Institute, Block EN, Sector V, Salt Lake, Kolkata 700 091, India 2Department of Mathematics, EM 4/1, Techno India Salt Lake (TISL), Sector V, Kolkata 700 091, India Quantum inseparabilities can be classified into three inequivalent forms: entanglement, Einstein-Podolsky- Rosen (EPR) steering, and Bell’s nonlocality. Bell-nonlocal states form a strict subset of EPR steerable states which also form a strict subset of entangled states. Recently, EPR steerable states are shown to be fundamental resources for one-sided device-independent quantum information processing tasks and, hence, identification of EPR steerable states becomes important from foundational as well as informational theoretic perspectives. In the present study we propose a new criteria to detect whether a given two-qubit state is EPR steerable. From an arbitrary given two-qubit state, another two-qubit state is constructed in such a way that the given state is EPR steerable if the new constructed state is entangled. Hence, EPR steerability of an arbitrary two-qubit state can be detected by detecting entanglement of the newly constructed state. Apart from providing a distinctive way to detect EPR steering without using any steering inequality, the novel finding in the present study paves a new direction to avoid locality loophole in EPR steering tests and to reduce the “complexity cost” present in experimentally detecting EPR steering. We also generalise our criteria to detect EPR steering of higher dimensional quantum states. Finally, we illustrate our result by using our proposed technique to detect EPR steerability of various families of mixed states.

I. INTRODUCTION ing EPR steering which also form a strict subset of entangled states [9]. One important point to be stressed here is that EPR Einstein-Podolsky-Rosen (EPR) steering is defined as the steering is inherently asymmetric with respect to the observers apparent ability to affect a spatially separated quantum state, unlike quantum nonlocality and entanglement [10]. In this which was the central problem in the EPR argument [1] to case, the outcome statistics of one subsystem (which is be- demonstrate the incompleteness of quantum mechanics. In ing ‘steered’) is produced due to quantum measurements on particular, EPR argument considers an entangled state shared a quantum state. However, there is no such constraint for between two spatially separated parties and it implies the the other subsystem. In fact it is shown that there exist en- possibility to produce different set of states at one party’s tangled states which are one-way steerable, i.e., demonstrate end by performing local quantum measurements of any two steerability from one observer to the other spatially separated non-commuting observables on another spatially separated observer, but not vice-versa [10–12]. Apart from having foun- party’s end. This “Spooky action at a distance” motivated dational significance, EPR steering has a vast range of infor- Schrodinger¨ to conceive the celebrated concept of ‘EPR steer- mation theoretic application in one-sided device-independent ing’ [2]. However, the research field of quantum steering did scenario where the party, which is being steered, has trust not progress much until 2007, when Wiseman, Jones, and Do- on his/her quantum device but the other party’s device is herty (WJD) introduced the concept of EPR steering in the untrusted. These applications range from one-sided device- form of a task [3,4]. The task of quantum steering is that independent quantum key distribution [13], advantages in a referee has to determine (using the measurement outcomes sub-channel discrimination [14], secure quantum teleporta- communicated classically from the two parties to the referee) tion [15, 16], quantum communication [15], detecting bound whether two spatially separated parties share entanglement, entanglement [17], one-sided device-independent randomness when one of the two parties is untrusted. WJD introduced generation [18], one-sided device-independent self-testing of the notion of EPR steering as the inability to construct a local pure maximally as well as non-maximally entangled states hidden variable-local hidden state (LHV-LHS) model to ex- [19]. plain the joint probabilities of measurement outcomes. Note arXiv:1808.07848v2 [quant-ph] 14 May 2019 Against this above backdrop, from fundamental viewpoint that in EPR steering scenario the no-signalling condition (the as well as from information theoretic perspective it is impor- probability of obtaining one party’s outcome does not depend tant to detect EPR steerable states. A number of criteria to on spatially separated other party’s measurement setting) is detect EPR steering have been proposed till date [20–30]. In always satisfied by the spatially separated two parties. the present study we provide a completely new criteria to de- It is well-known that EPR steering is an intermediate form tect EPR steering of an arbitrary two-qubit state. Given an of quantum inseparabilities in between entanglement [5] and arbitrary two-qubit state, a new two-qubit state is constructed Bell nonlocality [6–8]. Quantum states that demonstrate Bell- in such a way that EPR-steering of the given state is detected nonlocality form a strict subset of quantum states demonstrat- if the new constructed state is entangled. Hence, following Peres-Horodecki criteria [31, 32] we can state that the given state is EPR-steerable if the partial transpose of the new con- ∗ [email protected] structed state has at least one negative eigenvalue. We present † [email protected] a brief idea on how our result can be implemented in experi- ‡ [email protected] ment. A possible extension of our result to the case of higher 2 dimensional quantum systems has also been demonstrated. forms arbitrary measurement A with measurement operators A A P A The present study indicates a connection between EPR Ma (Ma ≥ 0 ∀ A, a and a Ma = I ∀ A) corresponding to the steering and entanglement. The novelty of the result ob- outcome a, Bob’s (unnormalized) conditional state becomes tained in the present study is that it presents a method to σA = Tr [(MA ⊗ I) ρ ], (1) detect EPR steering without using any steering inequality. a A a AB From experimental point of view, the present study enables where I is the 2×2 identity matrix. On this unnormalized con- one to indirectly test EPR steering of an arbitrary two-qubit ditional state Bob performs measurement B with measurement state through entanglement witness [33–38] of the new con- B B operators Mb corresponding to the outcome b (Mb ≥ 0 ∀ B, b structed state. Hence, our proposed theorem enables to re- P B and b Mb = I ∀ B) to produce the joint probability distri- duce the “complexity cost” [39] (it quantifies how complex B A bution P(a, b|A, B, ρAB) = Tr[Mb σa ], where P(a, b|A, B, ρAB) an experiment is in order to determine entanglement, EPR denotes the joint probability of obtaining the outcomes a and steering, Bell nonlocality) in experimentally determining EPR b, when measurements A and B are performed by Alice and steering as it has been shown that the “complexity cost” for Bob locally on state ρAB, respectively. the least complex demonstration of entanglement is less than The bipartite state ρAB of the system is steerable from Alice the “complexity cost” for the least complex demonstration of to Bob if and only if it is not the case that for all A ∈ Fα, EPR steering [39]. In particular, new nonlocality tests were B ∈ Fβ, a ∈ Ga, b ∈ Gb, the joint probability distribution can constructed to demonstrate the complexity costs of entangle- be written in the form ment and EPR steering. These inequalities are the simplest X possible witnesses for the above two types of quantum in- P(a, b|A, B, ρAB) = P(λ) P(a|A, λ) PQ(b|B, ρλ), (2) separabilities. On the other hand, Bell-CHSH (Bell-Clauser- λ Horne-Shimony-Holt) inequality is the simplest possible wit- where P(λ) is the probability distribution over the local hidden ness for Bell nonlocality. The above complexity costs have P variables (LHV) λ with λ P(λ) = 1. P(a|A, λ) denotes an ar- also been demonstrated experimentally by showing the vio- B bitrary probability distribution and PQ(b|B, ρλ)(= Tr[ρλ Mb ]) lations of these new inequalities and Bell-CHSH inequality denotes the quantum probability of outcome b given measure- using photonic singlet states [39]. B ment B on the local hidden state (LHS) ρλ; Mb being the mea- The present study may be helpful to avoid the locality loop- surement operator of the observable B associated with out- hole present in EPR steering test. Because the degree of corre- come b. In other words, the bipartite state ρAB is steerable lation required for entanglement testing is smaller than that for from Alice to Bob if and only if it does not have LHV-LHS violation of a steering inequality, it should be correspondingly model description (2) for arbitrary measurements performed easier to demonstrate entanglement without making the fair- by Alice and Bob. sampling assumption [40]. Our proposed procedure to test A bipartite state ρAB is called separable if and only if the EPR steering through entanglement detection makes exper- state can be written in the following form imental demonstration of EPR steering easier since demon- X A B strating EPR steering is strictly harder than demonstrating en- ρAB = P(λ) ρλ ⊗ ρλ , (3) tanglement as mentioned in Ref. [40]. One important point λ to be stressed here is that Bell nonlocality can be indirectly where P P(λ) = 1. A bipartite state, which is not separable, detected by detecting EPR steering [41]. The present study, λ is called entangled. Alternatively, the bipartite state ρAB of the therefore, completes demonstrating the connections between system is entangled if and only if it is not the case that for three inequivalent forms of quantum inseparabilities. all A ∈ Fα, B ∈ Fβ, a ∈ Ga, b ∈ Gb, the joint probability We organize this paper in the following way. We briefly distribution can be written in the form discuss the concept of EPR steering and entanglement in Sec- X A B tionII. In Section III, we present the main result of this pa- P(a, b|A, B, ρAB) = P(λ) PQ(a|A, ρλ ) PQ(b|B, ρλ ), (4) per on detecting EPR steering of an arbitrary two-qubit state λ indirectly through entanglement detection. Whether this re- P A A A where λ P(λ) = 1. PQ(a|A, ρλ )(= tr[ρλ Ma ]) denotes the sult can be extended to higher dimensional system is also quantum probability of outcome a given measurement A on discussed in Section III. We illustrate our result by detecting A A the local hidden state ρλ ; Ma being the measurement operator EPR steerability of various classes of two-qubit mixed states of the observable A associated with outcome a. P (b|B, ρB) and qubit-qutrit mixed states using our proposed technique in Q λ is similarly defined. In other words, the bipartite state ρAB is SectionIV. Finally, in SectionV we summarize the results entangled if and only if it does not have LHS-LHS model de- obtained and present the concluding remarks. scription (4) for arbitrary measurements performed by Alice and Bob.

II. PRELIMINARIES III. DETECTING EPR STEERING THROUGH ENTANGLEMENT DETECTION Suppose A ∈ Fα and B ∈ Fβ are the possible choices of measurements for two spatially separated observers, say Alice and Bob, with outcomes a ∈ Ga and b ∈ Gb, respectively. Let the The main result of this paper is stated in the following the- state ρAB is shared between Alice and Bob. After Alice per- orem. 3

Theorem 1. For any two-qubit state ρAB shared between Alice P(a, +|A, zˆ, ρAB) is the joint probability of obtaining the out- 1 2 and Bob, deﬁne two new states τAB and τAB given by, comes a and +1, when measurement A and projective measurement corresponding to the operator σz are performed by 1 ˜1 τAB = µ1 ρAB + (1 − µ1) ρ AB, (5) Alice and Bob locally on state ρAB, respectively. and h zˆ Ai 2 ˜2 n4 = Tr π−ςa τAB = µ2 ρAB + (1 − µ2) ρ AB, (6) 1 ˜1 ⊗ I 1 = µ1P(a, −|A, zˆ, ρAB) + (1 − µ1)P(a|A, ρAB) , (11) where ρ AB = ρA 2 with ρA = TrB[ρAB] = TrB[τAB] being 2 the reduced state at Alice’s side; ρ˜2 = I ⊗ ρ with ρ = AB 2 B B zˆ 2 where π− is the projector onto the eigenstate of σz correspond- TrA[ρAB] = TrA[τ ] being the reduced state at Bob’s side; AB ing to the eigenvalue −1 and it is given by, µ ∈ [0, √1 ]; µ ∈ [0, √1 ]. If τ1 is entangled, then ρ is 1 3 2 3 AB AB 2 ! EPR steerable from Bob to Alice. On the other hand, if τAB is zˆ 0 0 π− = . (12) entangled, then ρAB is EPR steerable from Alice to Bob. 0 1 Proof. At ﬁrst we shall prove that if τ1 is entangled, then AB P(a, −|A, zˆ, ρ ) is the joint probability of obtaining the out- ρ is EPR steerable from Bob to Alice. We shall prove this AB AB comes a and −1, when measurement A and projective mea- by proving its converse negative proposition: if ρ is not EPR AB surement corresponding to the operator σ are performed by steerable from Bob to Alice, then τ1 is separable. z AB Alice and Bob locally on state ρ , respectively. Hence, from Let us calculate P(a, b|A, B, τ1 ) = Tr[MBςA], which is the AB AB b a Eqs.(9) and (11) we get joint probability of obtaining the outcomes a and b, when arbitrary measurements A and B are performed by Alice and n + n = P(a|A, ρ ) (13) 1 A 1 4 AB Bob locally on state τAB, respectively, where ςa is the (unnormalized) conditional state on Bob’s side when Alice performs and A measurement A with measurement operators Ma corresponding to the outcome a. n1 − n4 = 2µ1P(a, +|A, zˆ, ρAB) − µ1P(a|A, ρAB). (14)

A h A 1 i ςa = TrA Ma ⊗ I τAB h i A ˜1 h i 1 = TrA Ma ⊗ I µ1ρAB + (1 − µ1)ρ AB xˆ A Re[n2] = Tr π ς − P(a|A, ρAB) h i h i + a 2 A A ˜1 = µ1 TrA Ma ⊗ I ρAB + (1 − µ1) TrA Ma ⊗ I ρ AB µ1 = µ1P(a, +|A, xˆ, ρAB) − P(a|A, ρAB), (15) h A i I 2 = µ1 TrA Ma ⊗ I ρAB + (1 − µ1)P(a|A, ρAB) , (7) 2 xˆ where π+ is the projector onto the eigenstate of σx correspond- where P(a|A, ρAB) denotes the marginal probability of Alice ing to the eigenvalue +1 and it is given by, to obtain the outcome a contingent upon performing measure- 1 1 ment A on the state ρAB.     Let us assume that xˆ 2 2 π+ =   . (16) ! 1 1 A n1 n2   ςa = 2 2 n3 n4 n + n n − n P(a, +|A, xˆ, ρ ) is the joint probability of obtaining the out- = 1 4 I + Re[n ]σ − Im[n ]σ + 1 4 σ , (8) AB 2 2 x 2 y 2 z comes a and +1, when measurement A and projective measurement corresponding to the operator σx are performed by where n1, n4 are real and n2 = n¯3 withn ¯3 being the com- A Alice and Bob locally on state ρAB, respectively. plex conjugate of n3 (since ςa is a Hermitian matrix). σx, σy, σz are Pauli matrices. Re[n2] and Im[n2] denote the real and imaginary part of n2, respectively. h yˆ Ai 1 A Im[n2] = − Tr π+ςa + P(a|A, ρAB) Now let us evaluate the matrix elements of ςa using Eq. (7). 2 µ1 h zˆ Ai = −µ1P(a, +|A, yˆ, ρAB) + P(a|A, ρAB), (17) n1 = Tr π+ςa 2

1 yˆ = µ1P(a, +|A, zˆ, ρAB) + (1 − µ1)P(a|A, ρAB) , (9) where π+ is the projector onto the eigenstate of σy correspond- 2 ing to the eigenvalue +1 and it is given by where πzˆ is the projector onto the eigenstate of σ correspond- + z 1 −i ing to the eigenvalue +1 and it is given by,     yˆ 2 2  ! π+ =   . (18) zˆ 1 0  i 1  π+ = . (10)   0 0 2 2 4

P(a, +|A, yˆ, ρAB) is the joint probability of obtaining the out- Using the above Eqs. (20), (22), (23), (24) and (25) we get comes a and +1, when measurement A and projective mea- " ( surement corresponding to the operator σ are performed by X I y P(a, b|A, B, τ1 ) = Tr MB P(λ)P (a|A, ρA) Alice and Bob locally on state ρ , respectively. AB b Q λ 2 AB λ Combining Eqs. (8), (13), (14), (15), (17) we obtain X A + µ1 P(λ)PQ(a|A, ρλ )P(+|xˆ, λ) σx λ X A I µ1 A ς = P(a|A, ρ ) − P(λ)PQ(a|A, ρ ) σx a AB 2 λ 2 λ µ1 X + µ1P(a, +|A, xˆ, ρAB) − P(a|A, ρAB) σx A 2 + µ1 P(λ)PQ(a|A, ρλ )P(+|yˆ, λ) σy µ1 λ + µ1P(a, +|A, yˆ, ρAB) − P(a|A, ρAB) σy X 2 µ1 A − P(λ)PQ(a|A, ρλ ) σy µ1 2 + µ1P(a, +|A, zˆ, ρAB) − P(a|A, ρAB) σz, (19) λ 2 X A + µ1 P(λ)PQ(a|A, ρλ )P(+|zˆ, λ) σz λ Therefore, from Eq.(19) we get )# µ X − 1 P(λ)P (a|A, ρA) σ . (26) 2 Q λ z 1 B A λ P(a, b|A, B, τAB) = Tr[Mb ςa ] " n I Now let us choose = Tr MB P(a|A, ρ ) b AB 2 I + σ~ · ~rλ µ ρB = , (27) + µ P(a, +|A, xˆ, ρ )σ − 1 P(a|A, ρ )σ λ 2 1 AB x 2 AB x µ + µ P(a, +|A, yˆ, ρ )σ − 1 P(a|A, ρ )σ where σ~ = (σx, σy, σz) is a vector composed of Pauli matrices 1 AB y 2 AB y and # µ1 o + µ1P(a, +|A, zˆ, ρAB)σz − P(a|A, ρAB)σz . ~r = µ 2P(+|xˆ, λ) − 1, 2P(+|yˆ, λ) − 1, 2P(+|zˆ, λ) − 1 . (28) 2 λ 1 (20) To be a valid probability distribution, P(+|xˆ, λ), P(+|yˆ, λ), P(+|zˆ, λ) ∈ [0, 1]. It can be easily checked that |~r | ≤ 1 im- " # λ Now, if ρAB is not steerable from Bob to Alice, then for all A ∈ 1 B plies µ1 ∈ 0, √ . Therefore, ρ is a valid quantum state Fα, B ∈ Fβ, a ∈ Ga, b ∈ Gb, the joint probability distribution 3 λ " # can be written in the form 1 (qubit) for µ1 ∈ 0, √ . 3 X A P(a, b|A, B, ρAB) = P(λ)PQ(a|A, ρλ )P(b|B, λ). (21) From the above construction one can write down the fol- λ lowing for all A ∈ Fα, B ∈ Fβ, a ∈ Ga, b ∈ Gb,

Comparing Eqs. (26) and (29), we can write X A P(a, +|A, zˆ, ρAB) = P(λ)PQ(a|A, ρ )P(+|zˆ, λ), (24) λ 1 X A B λ P(a, b|A, B, τAB) = P(λ)PQ(a|A, ρλ )PQ(b|B, ρλ ). (30) λ and Hence, we can conclude that if for arbitrary measurement A ∈ F performed by Alice and for arbitrary measurement B ∈ F X α β A performed by Bob the joint probability distribution obtained P(a|A, ρAB) = P(λ)PQ(a|A, ρλ ). (25) λ from the state ρAB can be written in the form given by Eq.(21), 5 then the joint probability distribution obtained from the state However, partial transposition is a non-positive map as 1 τAB can always be written in the form given by Eq.(30). In transposition is not a completely positive map. Hence, partial other words, if ρAB is not EPR steerable from Bob to Alice, transposition is a non-physical operation. Therefore, one can- 1 then τAB is separable. not directly implement Observation1 in the laboratory. How- In a similar way, it can be shown that if ρAB is not EPR ever, Horodecki and Ekert proposed a method called “struc- 2 steerable from Alice to Bob, then τAB is separable. This com- tural physical approximation” by which non-physical opera- pletes the proof. tions such as partial transposition can be systematically ap- proximated by physical operations [43]. Interestingly, struc- Now we are going to present a brief outline of the possi- tural physical approximations to the non-physical operations ble experimental implementation of Theorem1. Suppose we (including partial transposition) can be factorized into local want to check experimentally whether an arbitrary given two- operations and classical communication [44]. Following the qubit state ρAB (shared between Alice and Bob) demonstrates structural physical approximation to partial transposition [43], EPR steering from Bob to Alice using Theorem1. In order to a two-qubit state τAB is entangled if and only if the smallest achieve this, at first one of the two qubits (say, Bob’s qubit) of 2 eigenvalue of the stateτ ÃB is less than 9 , where the given system with state ρAB is subjected to a local depolarizing channel. Hence, the state after the channel action is 2 1 τ˜ = Λ(τ ) = I ⊗ I + [I ⊗ T](τ ). (33) given by, AB AB 9 9 AB 3 f X † Here T is the transposition operation and partial transpose ρ = (I ⊗ Ki) ρAB (I ⊗ K ). (31) AB i of τAB is denoted by [I ⊗ T](τAB). Note that the above map i=0 Λ(τAB) = τÃB is a completely positive map and therefore physically implementable. This map can be implemented by Here Kis are the Kraus√ operators associated√ with depolarizing√ 1+3p 1−p 1−p applying selected products of unitary (Pauli) transformations channel with K = I, K = σ , K = σ , √ 0 2 1 2 x 2 2 y with certain probabilities [43]. The experimental demonstra- 1−p tion of the structural physical approximation to the partial K3 = 2 σz and p is the channel strength with 0 ≤ p ≤ 1. It can be easily checked that transpose of two-qubit system has been performed in photonic systems using linear optical devices [45]. Hence, one ρ f = p ρ + (1 − p)ρ ˜ , (32) can modify observation1 for experimental implication in the AB AB AB following way: whereρ ˜ = ρ ⊗ I with ρ = Tr [ρ ] being the reduced AB A 2 A B AB Observation 2. If the smallest eigenvalue of the state τ˜1 = state of the Alice’s qubit. AB 2 I⊗I + 1 [I⊗T](τ1 ) is less than 2 , then ρ is EPR steerable According to Theorem1, the given state ρ is EPR steer- 9 9 AB 9 AB AB from Bob to Alice. On the other hand, if the smallest eigen- able from Bob to Alice if the state ρ f given by Eq.(32) (with AB value of the state τ˜2 = 2 I ⊗ I + 1 [I ⊗ T](τ2 ) is less than 2 , 1 AB 9 9 AB 9 p ∈ [0, √ ]) is entangled. 1 3 then ρAB is EPR steerable from Alice to Bob; where ρAB, τAB, f 2 1 One can experimentally construct the state ρAB from the τAB are defined in the statement of Theorem1; [I ⊗ T](τAB) 2 1 2 given two-qubit state ρAB by implementing the depolarizing and [I ⊗ T](τAB) denote the partial transpose of τAB and τAB channel following the proposed techniques (see [42] and the respectively. references therein) and then subjecting Bob’s qubit to the local depolarizing channel. Finally, entanglement of the state Hence, in order to experimentally check EPR steering of a f given two-qubit state ρAB, one has to experimentally prepare ρAB can be experimentally tested using entanglement witness 1 2 ([37, 38] and the references therein). the states τAB and τAB from the given state ρAB by applying local depolarizing channel as described earlier. Then the states Similarly, one can experimentally check EPR steering of 1 2 τÃB andτ ÃB described above can be experimentally prepared the given two-qubit state ρAB from Alice to Bob by subjecting 1 2 Alice’s qubit to a local depolarizing channel, and then check- from τAB and τAB, respectively, using local measurements and classical communication following the technique adopted in ing entanglement of the constructed state using entanglement 1 [43, 45]. Finally, the minimum eigenvalue of the statesτ ÃB witness. 2 According to the famous Peres-Horodecki criteria [31, 32], andτ ÃB can be estimated by defining a classical optimization problem over the measurement outcomes [45] to determine any given two-qubit state is entangled if and only if the partial 1 2 transpose of the given state has at least one negative eigen- whether the statesτ ÃB andτ ÃB are entangled. value. Hence, Theorem1 immediately provides the following It is to be noted that the criteria stated in Theorem1 to important observation. detect EPR steering through entanglement detection is applicable when the shared state belongs to the Hilbert space 1 2 2 Observation 1. If the partial transpose of the state τAB has at H ⊗ H . However, Theorem1 can be generalized to the case d 2 least one negative eigenvalue, then ρAB is EPR steerable from when the shared state belongs to the Hilbert space H ⊗ H , Bob to Alice. On the other hand, if the partial transpose of where d is arbitrary. We have proved in Theorem1 that if ρAB 2 1 the state τAB has at least one negative eigenvalue, then ρAB is not EPR steerable from Bob to Alice, then τAB is separa- 1 2 is EPR steerable from Alice to Bob; where ρAB, τAB, τAB are ble. However, the proof of Theorem1 remains unchanged if A defined in the statement of Theorem1. the dimension of the system at Alice’s end is d. Note that Ma 6 stated in the proof of Theorem1 is the measurement operator as mixed) by detecting entanglement of another constructed corresponding to the outcome a when Alice performs mea- two-qubit state. In the following we will detect EPR steer- surement A on her “qubit”. The whole mathematical proof ing of different families of two-qubit mixed states using our of Theorem1 remains unchanged if the above measurement proposed Theorem1. These families of states are chosen for operator is assumed to be acted on Alice’s “qudit”. Thus the various reasons. Some families are chosen as their experi- proof is independent of the dimension of the system at Alice’s mental preparations have been reported. Some other families end. Hence, we can state the following observation. of states are chosen in order to compare Theorem1 with the results obtained by using Semi-definite Program (SDP). Fi- Observation 3. For any qudit-qubit state ρAB shared between 1 nally, we will use Observation3 to detect EPR steering of a Alice and Bob, define another new qudit-qubit state τAB given family of qubit-qutrit mixed states. by • Consider that the following two-qubit Werner state is 1 ˜1 shared between Alice and Bob. τAB = µ1 ρAB + (1 − µ1) ρ AB, (34) I I where ρ˜1 = ρ ⊗ I with ρ = Tr [ρ ] = Tr [τ1 ] being ρAB = p |ψihψ| + (1 − p) ⊗ , (36) AB A 2 A B AB B AB 2 2 the reduced state (qudit) at Alice’s side; µ ∈ [0, √1 ]. If τ1 1 3 AB where |ψi = √1 (|01i − |10i) is the singlet state, {|0i, |1i} is entangled, then ρAB is EPR steerable from Bob to Alice. On 2 2 the other hand, for any qubit-qudit state ρAB shared between being the orthonormal basis in C , 0 ≤ p ≤ 1. Experimen- 2 Alice and Bob, define another new qubit-qudit state τAB given tal preparation of Werner state via spontaneous parametric by down-conversion and controlled decoherence of photons was demonstrated in [46]. We want to test in which range of p the 2 ˜2 τ = µ ρAB + (1 − µ ) ρ , (35) AB 2 2 AB state ρAB given by Eq.(36) is steerable from Bob to Alice or that from Alice to Bob using Theorem1. ˜2 I 2 where ρ AB = ⊗ ρB with ρB = TrA[ρAB] = TrA[τ ] being 2 AB Following Theorem1, from the given two-qubit state ρAB the reduced state (qudit) at Bob’s side; µ ∈ [0, √1 ]. If τ2 is 2 3 AB (36) we construct the new two-qubit state given by entangled, then ρAB is EPR steerable from Alice to Bob. τ1 = µ ρ + (1 − µ ) ρ˜1 , (37) Note that Observation3 enables to detect EPR steering AB 1 AB 1 AB from Bob to Alice of a qudit-qubit state shared between Alice where ρ˜1 = ρ ⊗ I with ρ = Tr [ρ ] = I being the and Bob, but it does not enable to detect EPR steering from AB A 2 A B AB 2 reduced state at Alice’s side. We choose µ = √1 . The newly Alice to Bob of the qudit-qubit state shared between Alice and 1 3 1 Bob. Similarly, using Observation3 one can detect EPR steer- constructed state τAB given by Eq. (37) is entangled for p > ing from Alice to Bob of a qubit-qudit state shared between √1 which can be checked using Peres-Horodecki criteria [31, 3 Alice and Bob, but can not detect EPR steering from Bob to 32]. Hence, Theorem1 concludes that the given two-qubit Alice of the qubit-qudit state shared between Alice and Bob. Werner state ρAB (36) is steerable from Bob to Alice for p > One important point to be stressed here is that the proof √1 . 3 of Theorem1 is based on the fact that the unnormalized con- In a similar way, using Theorem1 one can show that the A × ditional state on Bob’s side ςa is a 2 2 matrix mentioned given two-qubit Werner state ρ (36) is steerable from Alice in Eq.(8). Hence, this proof is only applicable when Bob’s AB to Bob for p > √1 . system is a qubit. That is why the mathematical procedure 3 described in the proof of Theorem1 cannot be generalized to Note that using quantum violation of 2-settings linear steer- detect EPR steering of shared state belonging to the Hilbert ing inequality proposed in [21] (we have used the particular d1 d2 form of this inequality mentioned in [28]), Werner state (36) space H ⊗ H (where, d1 and d2 are arbitrary). Further research is needed to investigate whether EPR steering of an is steerable (from Alice to Bob and from Bob to Alice) for p > √1 . On the other hand, Werner state is both-way steer- arbitrary dimensional state can be detected through entangle- 2 able for p > √1 using 3-settings linear steering inequality ment detection using different procedure. 3 [21, 28]. Hence, in this case our proposed technique to check EPR steerability through entanglement detection provides ad- IV. ILLUSTRATION WITH EXAMPLES vantage with respect to 2-settings linear steering inequality. However, our proposed technique and 3-settings linear steer- Any two-qubit pure entangled state is EPR-steerable [25]. ing inequality detect steerability of Werner state in the same The detection of entanglement of any given two-qubit pure region. Unsteerability of Werner state has been demonstrated 1 state thus certifies the EPR steerability of the given state. by constructing LHS model using SDP in the region p ≤ 2 However, this is not true for arbitrary mixed two-qubit states, [30]. i. e., there exists two-qubit mixed entangled states, which are One important point to be stressed here is that the converse 1 unsteerable. Hence, for arbitrary two-qubit mixed states EPR of Theorem1 is not always true, i. e., if τAB is separable, then steering cannot be detected by detecting entanglement of that ρAB may or may not be EPR steerable from Bob to Alice. On 2 state. The novelty of Theorem1 is that it enables to detect the other hand, if τAB is separable, then ρAB may or may not EPR steering of any arbitrary two-qubit state (pure as well be EPR steerable from Alice to Bob. From the above example 7 it can be checked using Peres-Horodecki criteria [31, 32] that the newly constructed state τ1 (37) (with µ = √1 ) is sepa- AB 1 3 rable for p ≤ √1 . However, the given two-qubit Werner state 3 1 ρAB (36) is steerable for p > 2 [3, 30]. Hence, in the region 1 < p ≤ √1 the newly constructed two-qubit state τ1 (37) is 2 3 AB separable, but the given two-qubit state ρAB (36) is steerable. • We will now check EPR steerability of a class of maximally entangled mixed state (MEMS) proposed by Munro et al. [47]. MEMS are those states that achieve the greatest possible entanglement for a given mixedness. The MEMS proposed by Munro et al. is given by    C  h(C) 0 0   2     0 1 − 2h(C) 0 0    ρmunro =   , (38)  0 0 0 0       C   0 0 h(C) 2 FIG. 1: Black region is denoted by ‘Region 1’. The region where with small square is denoted by ‘Region 2’. The region with  large square is denoted by ‘Region 3’. Theorem1 detects 1 2 EPR steerability of the Werner derivative state given by Eq.  , if C < 3 3 (40) for the values of state parameters α and θ indicated by h(C) =  , (39) C 2 Region 1 or Region 2 or Region 3. Quantum violation of  , if C ≥ 2 3 3-settings linear steering inequality [21, 28] detects EPR steerability of the Werner derivative state (40) for the values with C denoting the concurrence of the state ρ (38). Ex- munro of α and θ indicated by Region 2 or Region 3, but not by perimental technique to prepare this state using correlated Region 1. Quantum violation of 2-settings linear steering photons from parametric down-conversion has been presented inequality [21, 28] detects EPR steerability of the Werner [48]. derivative state (40) for the values of α and θ indicated by Munro state ρ (38) demonstrates both-way EPR steer- munro Region 3, but not by Region 1 and Region 2. ability for C > 0.531 following Theorem1. On the other hand, Munro state demonstrates both-way steering for C > 0.707 and for C > 0.667 using quantum violations of 2-settings linear steering inequality and 3-settings linear steering inequality, respectively. Therefore, in the region 0.531 < C ≤ 0.667, steerability of Munro state ρmunro (38) can be detected using Theorem1, but not using 2-settings linear steering inequality and 3-settings linear steering inequality. for a larger region of α and θ than 2-settings linear steering in- • We now focus on a class of non-maximally entangled equality and 3-settings linear steering inequality. The unsteer- mixed states (NMEMS). The states, which are not MEMS, ability of the state ρwd given by Eq. (40) was analyzed using are called NMEMS. In particular, we investigate EPR steer- SDP in Ref. [30]. The range of the state parameters α and θ ing of the Werner derivative states [49] which can be obtained where Werner derivative state (40) demonstrates EPR steering by applying a nonlocal unitary operator on the Werner state. was also presented in [30] using 9 projective measurements Werner derivative state is given by and the SDP method of Refs. [23, 25]. By comparing the result presented in [30] with Fig.1, it can be checked that SDP I I technique adapted in [23, 25, 30] demonstrates EPR steering ρ = α |ψ ihψ | + (1 − α) ⊗ , (40) wd θ θ 2 2 of Werner derivative state (40) for a larger region of state parameters compared to the technique presented in Theorem1. π where |ψθi = cos θ |00i + sin θ |11i with 0 ≤ θ ≤ 4 and For example, from Fig.1 it is evident that Theorem1 cannot −1 0 ≤ α ≤ 1. The state ρwd is entangled for α > [1+2 sin(2θ)] . detect EPR steering of the state ρwd with α = 0.55 for any val- In Fig.1 we have shown the region of α and θ for which ues of θ. However, SDP technique adapted in [23, 25, 30] can both-way EPR steering of Werner derivative state (40) is de- detect EPR steering of the state ρwd with α = 0.55 for some tected using Theorem1, quantum violation of 2-settings linear values of θ. steering inequality and quantum violation of 3-settings linear steering inequality, respectively. From this Figure it is clear that Theorem1 detects steerability of Werner derivative state • Let us consider another class of NMEMS proposed in [50] 8 given by 2-settings linear steering inequality as well as 3-settings linear steering inequality detect both-way steerability of the state  p + 2   0 0 0  ρτ for −1 ≤ τ ≤ 1. Theorem1 proposed in this study detects  6  both-way steerability of the state ρ for −1 ≤ τ < −0.366   τ  1 − p 1 − p  and for 0.366 < τ ≤ 1. Hence, in this case 2-settings linear    0 0  steering inequality and 3-settings linear steering inequality are  3 3  ρp =   , (41) more useful in detecting EPR steerability than our proposed  1 − p 1 − p   0 0  theorem.  3 3    •  p Let us consider the following class of state, which demon-  0 0 0  strate one-way steerability [10], is shared between Alice and 2 Bob where 0 ≤ p ≤ 1. Note that the above state can be written as 1 − α I I G W ρα = α |ψihψ| + 2 |0ih0| ⊗ + 3 ⊗ |1ih1| , (47) ρp = p ρAB + (1 − p) ρAB, (42) 5 2 2

G | ih | W | ih | 1 where ρAB = TrC[ GHZ GHZ ] and ρAB = TrC[ W W ]. where |ψi = √ (|01i − |10i) and 0 ≤ α ≤ 1. 2-settings linear |GHZi and |Wi are the GHZ (Greenberger-Horne-Zeilinger) 2 steering inequality detects both-way EPR steerability of the state and W state given by, 1 above state ρα for α > √ = 0.707. 3-settings linear steering 2 1 inequality detects both-way EPR steerability of the above state | i | i | i GHZ = √ ( 000 + 111 ), (43) for α > √1 = 0.577. Our proposed Theorem1 detects EPR 2 3 steerability from Bob to Alice of the above state ρα for α > and 0.577. Interestingly, Theorem1 detects EPR steerability from 1 Alice to Bob of the state ρα (47) for α > 0.566. It is to be noted |Wi = √ (|001i + |010i + |100i). (44) that the state ρα (47) demonstrates one-way steering in the 3 1 range 0.4983 . α ≤ 2 [10]. In particular, it has been shown The state (41) has experimental relevance since it has been that the state ρα (47) does not demonstrate EPR steering from ≤ 1 constructed from two qubit GHZ and W states, and GHZ as Bob to Alice for α 2 . On the other hand, using SDP it well as W states are both experimentally realizable in photonic has been shown that the above state is steerable from Alice systems using spontaneous parametric down-conversion [51, to Bob for α & 0.4983. Hence, SDP used in Ref. [10] is 52]. more efficient than our proposed Theorem1 in detecting EPR steering of the state ρα from Alice to Bob. Theorem1 detects both-way EPR steerability of ρp given by Eq. (41) for p < 0.073. On the other hand, EPR steerabil- • Consider that an initial maximally entangled two-qubit 1 state |ψii = √ (|00i + |11i), shared between Alice and Bob, ity of the above state is not detected using 2-settings linear 2 steering inequality and 3-settings linear steering inequality. is subjected to independent local amplitude damping channel • P2 P2 † † Now we will study EPR steering of the following class given by the evolution ρ f = i=1 j=1 Ei ⊗ E j ρi Ei ⊗ E j . of maximally steerable mixed state (MSMS) (the states that Here ρi = |ψiihψi| is the density matrix of the initially shared violate to the most degree a steering inequality for a given state, ρ f is the density matrix of the final state. The two p mixedness) proposed by Ren et al. [53], | ih | − | ih | Kraus operators√ are defined as E0 = 0 0 + 1 p 1 1 1 − τ 1 − τ and E1 = p |0ih1| with 0 ≤ p ≤ 1. Note that experi-  0 0  mental technique has already been realized to engineer am-  4 4    plitude damping channel [54]. 2-settings linear steering in-  1 + τ 1 + τ  equality detects both-way EPR steerability of the final state  0 0   4 4  for 0 ≤ p < 0.293. 3-settings linear steering inequality detects ρτ =   , (45)  1 + τ 1 + τ  both-way EPR steerability of the final state for 0 ≤ p < 0.397.  0 0   4 4  Our proposed Theorem1 detects both-way EPR steerability   ≤ 1 − τ 1 − τ of the above final state for 0 p < 0.411. Hence, in this  0 0  case 2-settings linear steering inequality and 3-settings linear 4 4 steering inequality are less useful in detecting EPR steerabil- where −1 ≤ τ ≤ 1. Note that the above state can be written in ity than our proposed theorem. Importantly, it has been shown the following form [29] using SDP that the above final state has LHS model for p & 0.4. Hence, the approximate lower bound of p where the 1 − τ 1 + τ ρ = ρ + ρ , (46) final state has LHS model obtained using SDP in Ref. [29] τ 2 1 2 2 almost matches with the upper bound of p where our pro-

1 posed theorem detects EPR steering of the aforementioned fi- where ρi = |ψiihψi| with |ψ i = √ (|00i + |11i) and |ψ i = 1 2 2 nal state. 1 √ (|01i + |10i). Hence, the state ρτ can be prepared as a 2 • Let us focus on checking EPR-steerability of some higher probabilistic mixture of two Bell states. dimensional quantum system. In particular, let us consider the 9 following qubit-qutrit state shared between Alice and Bob, date [20–30]. In the present study we have provided a novel criteria to detect EPR steering of an arbitrary two-qubit state. I ρµ = (1 − µ)ρw + µ ⊗ |vihv|, (48) This criteria enables one to detect EPR steering of the given 2 two-qubit state by detecting entanglement of another con- I I structed two-qubit state. Hence, theoretically one can detect where 0 ≤ µ ≤ 1; ρw = p |ψihψ| + (1 − p) ⊗ is EPR steering of the given state using Peres-Horodecki criteria 2 2 the Werner state mentioned in Eq.(36) with 0 ≤ p ≤ 1; without using any steering inequality. Besides having foun- |ψi = √1 (|01i − |10i) is the singlet state. |vi is a vacuum dational importance, our proposed technique to detect EPR 2 state orthogonal to Bob’s qubit subspace. The above state (48) steering through entanglement detection reduces the “com- can be prepared when one subsystem of Werner state (36) is plexity cost” in experimentally determining EPR steering as distributed through a lossy channel to Bob. Note that a lossy the “complexity cost” for the least complex demonstration of channel is one that replaces a qubit with the vacuum state |vi entanglement is less than the “complexity cost” for the least with probability µ and can be represented by the map ρ → complex demonstration of EPR steering [39]. Moreover, this (1 − µ)ρ + µ|vihv|. We will now check EPR steerability of study may pave a new way in avoiding locality loophole in the qubit-qutrit state (48) following Observation3. Hence, we detecting EPR steering experimentally. It is to be noted that any quantum state, which is EPR steer- construct the following new qubit-qutrit state from the state ρµ mentioned in Eq. (48), able, is entangled as well, since quantum states demonstrating EPR steering form a strict subset of the entangled states [9]. 1 1 Hence, entanglement can be detected by detecting EPR steer- τ = √ ρµ + 1 − √ ρ˜µ, (49) 3 3 ing. Although, the converse is not true always as there exist entangled unsteerable states [9]. For this reason, in general, I whereρ ˜µ = 2 ⊗ ρB with ρB = TrA[ρµ] being the reduced state detecting entanglement in a quantum state does not guarantee (qutrit) of ρµ at Bob’s state. The qubit-qutrit state (49) is en- that the state under consideration is EPR steerable. However, tangled for √1 < p ≤ 1 for any values of µ. This can be the novelty of the present study is that it provides an indirect 3 checked through Peres-Horodecki criteria [31, 32]. Hence, way to detect EPR steering though entanglement detection. following Observation3 we can conclude that the qubit-qutrit Previously, it was shown that Bell nonlocality can be indi- state mentioned in Eq. (48) is EPR steerable from Alice to rectly detected by detecting EPR steering [41]. Hence, the Bob for √1 < p ≤ 1 for any values of µ. present study together with the results obtained in [41] con- 3 It has been shown [55, 56] that the qubit-qutrit state (48) is nect three inequivalent forms of quantum inseparabilities. 1 The present study is restricted to only qubit-qubit, qubit- EPR steerable from Alice to Bob for 2 < p ≤ 1 for any values of µ. Hence, the technique adopted in Refs. [55, 56] is more qudit and qudit-qubit systems. Whether EPR steering can be useful than Observation3 in detecting EPR steerability of the detected through entanglement detection in arbitrary higher qubit-qutrit state (48) from Alice to Bob. One important point dimensional system (qudit-qudit) as well as in multipartite to be stressed here is that the state (48) is one-way EPR steer- scenario is worth to be studied in future. In the present study 1 we have discussed some possible outlines to experimentally able since for 2 < p ≤ 1 the state (48) is not EPR steerable from Bob to Alice for certain values of µ considering projec- detect EPR steering through our proposed criteria using local tive measurements or positive operator valued measurements depolarizing channel together with entanglement witness (see, (POVM) [55, 56]. On the other hand, our proposed Obser- for example, [37, 38] and the references therein) or “struc- vation3 cannot detect EPR steering of the qubit-qutrit state tural physical approximation” to the partial transpose opera- (48) from Bob to Alice. Hence, we cannot use Observation3 tion [43, 45]. Hence, the experimental realization of our pro- to demonstrate one-way EPR steering of the qubit-qutrit state posed criteria following the above ideas is another area for (48). future research. As mentioned earlier, the experimental preparations of some of the aforementioned families of two-qubit states have been reported. Hence, after preparing these states following ACKNOWLEDGEMENTS the proposed techniques one can use local depolarizing channel [42] and structural physical approximation to partial trans- The authors acknowledge the anonymous referees for position [45] as mentioned in Section III in order to experi- valuable comments. D. D. and S. S. acknowledge fruitful mentally detect EPR steering of these families of states using discussions with Arkaprabha Ghosal, Som Kanjilal and our proposed Observation2. Arup Roy. D. D. acknowledges the financial support from University Grants Commission (UGC), Government of India. S. S. acknowledges the financial support from INSPIRE V. CONCLUDING DISCUSSIONS programme, Department of Science and Technology, Govern- ment of India. Since EPR steering has foundational significance as well as information theoretic applications, detecting EPR steering is Note added- After communicating this manuscript to the one of the most profound problem in recent times. A num- journal, a related work by Changbo Chen et al. [57] appeared ber of criteria to detect EPR steering has been proposed till in arXiv. 10

[1] A. Einstein, B. Podolsky, and N. Rosen, Phys. Rev. 47, 777 [28] A. C. S. Costa, and R. M. Angelo, Phys. Rev. A 93, 020103(R) (1935). (2016). [2] E. Schrodinger, Proc. Cambridge Philos. Soc. 31, 555 (1935); [29] D. Cavalcanti, L. Guerini, R. Rabelo, and P. Skrzypczyk, Phys. E. Schrodinger, Proc. Cambridge Philos. Soc. 32, 446 (1936). Rev. Lett. 117, 190401 (2016). [3] H. M. Wiseman, S. J. Jones, and A. C. Doherty, Phys. Rev. Lett. [30] F. Hirsch, M. T. Quintino, T. Vertesi, M. F. Pusey, and N. Brun- 98, 140402 (2007). ner, Phys. Rev. Lett. 117, 190402 (2016). [4] S. J. Jones, H. M. Wiseman, and A. C. Doherty, Phys. Rev. A. [31] A. Peres, Phys. Rev. Lett. 77, 1413 (1996). 76, 052116 (2007). [32] M. Horodecki, P. Horodecki, and R. Horodecki, Physics Letters [5] R. Horodecki, P. Horodecki, M. Horodecki, and K. Horodecki, A 223, 1 (1996). Rev. Mod. Phys 81, 865 (2009). [33] M. Lewenstein, B. Kraus, J. I. Cirac, and P. Horodecki, Phys. [6] J. S. Bell, Physics 1, 195 (1965). Rev. A 62, 052310 (2000). [7] J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, Phys. [34] D. Bruss, J. I. Cirac, P. Horodecki, F. Hulpke, B. Kraus, M. Rev. Lett. 23, 880 (1969). Lewenstein, A. Sanpera, J. Mod. Opt. 49, 1399-1418 (2002). [8] N. Brunner, D. Cavalcanti, S. Pironio, V. Scarani, and S. [35] O. Guhne, P. Hyllus, D. Bruss, A. Ekert, M. Lewenstein, C. Wehner, Rev. Mod. Phys. 86, 419 (2014). Macchiavello, and A. Sanpera, Phys. Rev. A 66, 062305 (2002). [9] M. T. Quintino, T. Vertesi, D. Cavalcanti, R. Augusiak, M. [36] O. Guhne, P. Hyllus, D. Bruss, A. Ekert, M. Lewenstein, C. Demianowicz, A. Acin, and N. Brunner, Phys. Rev. A 92, Macchiavello, and A. Sanpera, J. Mod. Opt. 50 (6-7), 1079- 032107 (2015). 1102, (2003). [10] J. Bowles, T. Vertesi, M. T. Quintino, and N. Brunner, Phys. [37] O. Guhne, and G. Toth, Physics Reports 474, 1 (2009). Rev. Lett. 112, 200402 (2014). [38] J. Dai, Y. L. Len, Y. S. Teo, B.-G. Englert, and L. A. Krivitsky, [11] J. L. Chen, X. J. Ye, C. F. Wu, H. Y. Su, A. Cabello, L. C. Kwek, Phys. Rev. Lett. 113, 170402 (2014). and C. H. Oh, Scientific Reports 3, 2143 (2013). [39] D. J. Saunders, M. S. Palsson, G. J. Pryde, A. J. Scott, S. M. [12] J. Bowles, F. Hirsch, M. T. Quintino, and N. Brunner, Phys. Barnett, and H. M. Wiseman, New J. Phys. 14, 113020 (2012). Rev. A 93, 022121 (2016). [40] D. J. Saunders, S. J. Jones, H. M. Wiseman, and G. J. Pryde, [13] C. Branciard, E. G. Cavalcanti, S. P. Walborn, V. Scarani, and Nature Physics 6, 845 (2010). H. M. Wiseman, Phys. Rev. A. 85, 010301(R) (2012). [41] J.-L. Chen, C. Ren, C. Chen, X.-J. Ye, and A. K. Pati, Scientific [14] M. Piani, and J. Watrous, Phys. Rev. Lett. 114, 060404 (2015). Reports 6, Article number: 39063 (2016). [15] M. D. Reid, Phys. Rev. A 88, 062338 (2013). [42] A. Chiuri, S. Giacomini, C. Macchiavello, and P. Mataloni, [16] Q. He, L. Rosales-Zarate, G. Adesso, and M. D. Reid, Phys. Phys. Rev. A 87, 022333 (2013). Rev. Lett. 115, 180502 (2015). [43] P. Horodecki and A. Ekert, Phys. Rev. Lett. 89, 127902 (2002). [17] T. Moroder, O. Gittsovich, M. Huber, and O. Guhne, Phys. Rev. [44] C. M. Alves, P. Horodecki, D. K. L. Oi, L. C. Kwek, and A. K. Lett. 113, 050404 (2014). Ekert, Phys. Rev. A 68, 032306 (2003). [18] Y. Zhi Law, L. Phuc Thinh, J.-D. Bancal, and V. Scarani, [45] H.-T. Lim, Y.-S. Kim, Y.-S. Ra, J. Bae, and Y.-H. Kim, Phys. J. Phys. A: Math. Theor. 47, 424028 (2014); E. Passaro, D. Rev. Lett. 107, 160401 (2011). Cavalcanti, P. Skrzypczyk, and A. Acin, New J. Phys. 17, [46] Y.-S. Zhang, Y.-F. Huang, C.-F. Li, and G.-C. Guo, Phys. Rev. 113010 (2015); P. Skrzypczyk, and D. Cavalcanti, Phys. Rev. A 66, 062315 (2002). Lett. 120, 260401 (2018); B. Coyle, M. J. Hoban, E. Kashefi, [47] W. J. Munro, D. F. V. James, A. G. White, and P. G. Kwiat, arXiv:1806.10565 [quant-ph] (2018). Phys. Rev. A 64, 030302(R) (2001). [19] I. Supic and M. J. Hoban, New J. Phys. 18, 075006 (2016); [48] N. A. Peters, J. B. Altepeter, D. A. Branning, E. R. Jeffrey, T.-C. A. Gheorghiu, P. Wallden, and E. Kashefi, New J. Phys. 19, Wei, and P. G. Kwiat, Phys. Rev. Lett. 92, 133601 (2004). 023043 (2017); S. Goswami, B. Bhattacharya, D. Das, S. Sas- [49] T. Hiroshima, and S. Ishizaka, Phys. Rev. A 62, 044302 (2000). mal, C. Jebaratnam, and A. S. Majumdar, Phys. Rev. A 98, [50] S. Adhikari, A. S. Majumdar, S. Roy, B. Ghosh and N. Nayak, 022311 (2018). Quant. Inf. Comp., 10, 0398 (2010). [20] M. D. Reid, Phys. Rev. A. 40, 913 (1989). [51] M. Eibl, N. Kiesel, M. Bourennane, C. Kurtsiefer, and H. We- [21] E. G. Cavalcanti, S. J. Jones, H. M. Wiseman, and M. D. Reid, infurter, Phys. Rev. Lett. 92, 077901 (2004). Phys. Rev. A 80, 032112 (2009). [52] H. Lu, L.-K. Chen, C. Liu, P. Xu, X.-C. Yao, L. Li, N.-L. Liu, [22] S. P. Walborn, A. Salles, R. M. Gomes, F. Toscano, and P. H. B. Zhao, Y.-A. Chen, J.-W. Pan, Nat. Photonics 8, 364 (2014). Souto Ribeiro, Phys. Rev. Lett. 106, 130402 (2011). [53] C. Ren, H.-Y. Su, H. Shi, and J. Chen, Phys. Rev. A 97, 032119 [23] M. F. Pusey, Phys. Rev. A 88, 032313 (2013). (2018). [24] T. Pramanik, M. Kaplan, and A. S. Majumdar, Phys. Rev. A 90, [54] W. McCutcheon, A. McMillan, J. G. Rarity, and M. S. Tame, 050305(R) (2014). New J. Phys. 20, 033019 (2017). [25] P. Skrzypczyk, M. Navascues, and D. Cavalcanti, Phys. Rev. [55] D. A. Evans, and H. M. Wiseman, Phys. Rev. A 90, 012114 Lett. 112, 180404 (2014). (2014). [26] E. G. Cavalcanti, C. J. Foster, M. Fuwa, and H. M. Wiseman,J. [56] S. Wollmann, N. Walk, A. J. Bennet, H. M. Wiseman, and G. J. Opt. Soc. Am. B 32, A74 (2015). Pryde, Phys. Rev. Lett. 116, 160403 (2016). [27] P. Girdhar and E. G. Cavalcanti, Phys. Rev. A 94, 032317 [57] C. Chen, C. Ren, X.-J. Ye, and J.-L. Chen, arXiv:1810.10234 (2016). [quant-ph] (2018).